Analysis of price increase and how to move forward for OC&Gym

A report by group B4

Summary

The objective of this report is to analyse the impact of the price change that OC&Gym implemented in 2016 and help them find strategies to increase revenue further. This was done by analysing KPIs such as subscriber numbers, churn rate, revenue and lifetime value. The deep-dive into different segments has provided key insights that were used to derive recommendations that tackle the problem of stagnating revenues. The introduction of new annual memberships, both a normal and off-peak version, will increase lifetime value of subscribers and therefore improve the revenue situation of OC&Gym. Furthermore, subscriber engagement and churn must become major focus points for the company due to the high constant churn rate of 20% and the fact that improving subscriber engagement will improve lifetime value through the average length of subscription.

Introduction

A price increase in 2016 across subscription types led to a less than anticipated increase in revenue for budget gym operator OC&Gym. We have been commissioned by the board to investigate the impact of the price rise, provide recommendations regarding new pricing strategies and to determine the reasons behind the slowing revenue and subscription numbers.

The monthly standard subscription price increased by 29% while monthly flexible subscription prices increased by 19% in 2016. This led to a 40% revenue increase for 2016, though this was not significantly higher than the 30% average price increase over the previous three years given these large price increases. The annual growth of revenues has since slowed, however, with OC&Gym now in line with the average industry revenue growth rate of 4.9% [reference 1].

We hypothesise that the price change has led to a different segmentation of our subscriber base. That is, we expect the price increase to affect each subscriber segments differently. We believe that the price change had a short-term effect on churn behaviour, with temporarily high churn rates, though we expect that this reverted to the mean in the long-term across subscription types and segments.

We remark that OC&Gym needs to increase its subscriber lifetime value (LTV) in order to address its revenue difficulties. We propose introducing measures to increase the average subscription length of our members, which would have a positive knock-on effect on LTV. Introducing an annual subscription offering could help accomplish that. Introducing an annual off-peak membership could help fill our gym at quieter hours and attract new members, all while positively impacting the LTV of our subscribers.

Technologies and techniques

Analysis for this project was done using standard python libraries which are all included in anaconda python distribution. Our team has used Pandas to load and process CSV files for subscriber, visitation and pricing data. In order to perform timeseries analysis we imported python datetime utilities, and we used numpy to handle both timestamps and periods. Most visualisations in this notebook were created using matplotlib or seaborn which acts a wrapper for matplotlib library and removes boilerplate code. Seaborn allowed us to create faceted plots, which give a better insight into data with many categorical variables.

To prepare relevant datasets we relied heavily on the use of filtering through either pandas loc method or numpy select. Moreover, to make plots reproducible and scalable we implemented our own functions to both segment data and generate plots. More information on custom functions is available in comments and code description sections.

To compute our subscription duration, we noted that the ‘End Date’ given in the dataset was the last month subscribers went to the gym for every single month and thus we used the ‘End Date’ in our analysis as the last effective month of subscriptions (code in Appendix).

We decided to plot all our graphs from January 2010 to January 2018 as to not distract the reader by effects at the start of OC&Gym and effects at the end of our data and allow him to focus on data from January 2010 to January 2018 with the emphasis on the price chance in January 2016.

Finally, we tagged cells which take longer time to run with %%time Jupyter magic command. Certain datasets will require 2-8 minutes to be prepared and about 4-6 GB of RAM in order to be stored in memory.

The following cells will go through importing the libraries we used for this report, importing the datasets and cleaning them, through standardising the date times and categorising the data to make segmentation and analysis easier. It will also include functions and creation of new datasets that will be used later in the results.

In the following cell we created a dataframe that joined visitation information to compute the weekly activity of a subscriber (how often he/she goes to the gym er week) and its type which depends on whether a subscriber goes most often at peak or off-peak time. This dataset will be used for our segmentation analysis.

Filtering the customer dataframe to only keep columns needed :

This next cell comprise of a collection of functions used for exploratory data analysis. They facilitate segmentation of customers across time, membership type and whether they have joined or left a gym.

This following cell is a set of modified functions from the previous block in order to deal with data comparisons by club. These functions exclude segmentation by subscription type and provide some additional metrics that are relevant only for club comparisons.

In this next cell we created a dataframe that will be club specific. In it, we will compute the average length of subscriptions for those who joined in the first year, the average length of subscriptions for those who joined after that first year, the difference in average lengths between the two groups and the percentage change in subscription lengths.

This next section is used to prepare series of dataframes used to compare and analyse club data. They rely on modified customer segmentation functions and provide detailed breakdowns of club related metrics by month and year.

In this next cell we created a dataframe that will be used to compute per months the number of subscribers, whether overall or per categories (flexible or standard, depending on a specific club). This will allow us to compute the revenue as well as growth rates (for our revenue or our subscriber evolution).

Results

1. What happened when the price was increased? What is the current situation?

1. A. Subscriber numbers

1. A. 1. Total Subscriber

1. A. 1. a) Subscriber base

As OC&Gym is a subscription business, where most if not all its revenue come from subscriptions sold, the number of active subscribers is a key metric for the business and will have a direct impact on revenue. As such, we plotted the number of subscribers and the growth rate of this number per month to understand the subscriber base of OC&Gym.

We see from the below graphs that our subscriber base has grown steadily since 2010, at around 2% per month. Subscriber numbers rebounded quickly in 2016 following the large number of cancellations that occurred in relation to the price increase. This price increase led to an approximate 15% fall in total subscribers, though this proved to be a temporary dip and numbers were soon restored to their previous levels. The subscriber base has experienced a recent slowdown in growth, raising fears that subscriber numbers may have plateaued under the current pricing system.

1. A. 1. b) New Subscribers

Subscriber numbers have increased throughout the period, displaying some seasonality with spikes at the beginning of each year as people sign up as a result of New Year’s resolution. Recent data suggests that the number of new subscribers may have peaked. As OC&Gym opened on average 5-6 gyms from 2009 to 2016, and that newly opened gyms tend to have high new subscriber growth rates, it can explain some of the volatility in these rates that we saw at the beginning of the 2010s. The average new subscriber rate was 2% over the period, though this rate currently stands at around 1%, excluding the usual January spike, which comes to show that OC&Gym has had a harder time recently finding new subscribers. This could come from the price increase, but also from other factors that we will explore later.

1. A. 1. c) Churn Rate

As we have seen in the previous section that we are continuing to gain subscribers, we also need to explore churn to understand our total number subscribers. Churn is also a vital part of any subscription business and it thus central to our analysis here.

We computed our churn rate as the number of subscribers leaving a certain month divided by the number of subscribers at the beginning of the month.

The churn rate has remained remarkably flat throughout the period on aggregate, except for the temporary increase related to the 2016 price rise. The churn rate is consistently about 20% per month. This overall trend shows that across all gyms, we are not losing more subscribers once the price change has taken into effect and that our overall churn has not been affected by this price change. We will go into more detail in a later part of this report into churn and specifically churn per club.

1. A. 2. Customer Segments

1. A. 2. a) Split of Customer Segments

We then decided to go into more depth in our total number of subscribers and look across different segment and their impact into the makeup of our current subscribers. We looked at the impact of the type of membership chosen, the type of subscriber (whether he/she goes more at peak or non-peak time), but also at the age groups, the genders, the affluence groups, the club groups and the average weekly visits to the gym (see appendix for the last 4 segments).

We can see from the below graphs that almost 90% of our subscribers are of the Standard subscription type, which could mean that most subscribers are not interested in short term subscriptions or are planning to stay long enough that they would not be impacted by the 3-month notice of the standard subscription.

We can also see that we labelled roughly 33% of our subscribers as being “off-peak subscribers”, meaning most of their visits to the gym came at the quieter, non-peak hours. Thus, there might be an opportunity to develop this segment, as moreover the gym still has lots of capacity at non-peak hours (see Number of visitors in the gym over 2009 - 2018 per hour in the Appendix). It also could imply that the gym is very crowded at peak hours, as most subscribers go at that time. We will explore later in this report if that could have an impact on our churn.

1. A. 2. b) Important Customer Segments that differ

We wanted to analyse the differences across segments in terms of subscriber join rates and churn rates. This analysis was carried out for all segments across both Standard and Flexible subscription types and can be found in the appendix. No clear patterns among segments were found, apart from the difference between Flexible and Standard membership.

We found that Flexible subscription types had higher subscriber join rates and churn rates. This makes intuitive sense - the appeal behind the Flexible subscription type is the ease at which members can cancel their contracts if their plans change.

These graphs were calculated so that we can see the contribution of each segment to total Standard/Flexible subscriber join rate and total churn rate. In other words, these graphs enable to us to visualise which groups drive the join and churn rates. It is imperative, therefore, to consider the contribution of each segment relative to how much of the subscriber segment it makes up. The makeup of subscriber segments can be found in the appendix. Patterns across segments can still be identified using this method.

1. A. 3. Gyms

OC&Gym has opened gyms consistently from 2009 to 2016 and has since slowed down their openings up to reaching their goal of one gym per district. As we have seen in previously, the regular opening of clubs has allowed for regular increases of new subscribers. Understanding the importance of individual clubs in the subscriber numbers is vital for OC&Gym as it will allow them to target low and high performing clubs better than a global stategy would.

The below chart shows us the number of active subscribers at each of the 32 locations. The range in members is relatively narrow. Each gym has more than 1000 and less than 2250 members.

1. A. 3. a) Analysis of all gyms

An analysis of the total subscriber growth rate of all gyms over 2018 revealed that there are large differences in growth rates across clubs. Clubs that have opened more recently tend to have higher growth rates due to their relatively lower bases. The below tables indicate that subscriber numbers fell in several gyms during 2018. This may offer a suggestion as to why the total subscriber growth rate seems to be stagnating, as older gyms find it more difficult to grow at the same rates they once did.

1. A. 3. b) Comparison of High-growth and Low-growth locations

We aimed to investigate differences between a well-performing gym and a poorly performing gym. We ranked gyms based on subscriber growth rate in 2018 and selected one gym with a high growth rate and one gym with a low growth rate, ensuring they had reasonably similar opening dates. In this analysis, we used Brent as our well-performing gym and Southwark as our poorly performing gym.

We noticed that monthly subscriber growth rates were largely in line with each other, with rates for Brent being slightly higher on average over the period. More interestingly though was the fact that churn rate and subscriber join rates were consistent significantly greater for Brent over the period. It is clear that the higher churn rate for Brent was being sufficiently offset by a high acquisition rate.

This motivated us to investigate the differences further by seeing if we could see any differences in patterns across segments. This full analysis can be found in the appendix. In conclusion, we did not find any significant difference in patterns between the gyms across any of the segments we looked at. An example of this analysis is given below, where frequency of gym visit was chosen as the segment. There is no clear difference in the patterns in either graph. This suggested to us that some external factor may be influencing the subscriber growth rates in these gyms. We posit that Southwark’s central location may play a part, as this area of London is likely to be a very saturated area in terms of gyms.

1. B. Revenue and Average revenue per subscriber

1. B. 1. Total Revenue and Average revenue per subscriber (APRU)

Total revenue, as well as revenue growth, is shown in the graphs below. We notice that the total revenue curve follows a similar path to the total subscriber's curve, though the drop as a result of the 2016 price rise is much less pronounced. Revenue increased rapidly again as subscriber growth stabilised following this price hike, suggesting that subscribers are not very price sensitive. This raises the possibility of further price increases and begs the question whether the price rise was large enough in order to satisfy the revenue target OC&Gym had. Revenue increased by 40% in 2016 after the price increase.

The graph below show that there is obviously a direct relationship between the number of subscribers and the revenue the company generates. This relationship can be captured by Average Revenue per User (ARPU). (This is an important metric for a subscription company such as OC&Gym.) We see the direct impact the price increase had on ARPU, increasing it from £23 to £29. ARPU has decreased slightly in recent years, as it depends on the amount of users of subscription types.

1. C. Life time value of subcribers

1. C. 1. Total subscribers : subscription length over time and life time value of subscribers

The graph below shows the average subscriber length of stay over time. Apart from the short-term drop caused by the price change, the average duration has remained remarkably stable around 5 months.

Turning length of stay, by multiplying it with ARPU, we get a similar picture. The difference occurs after the price change at the end of 2015. We can see that LTV has been increase from ~120 GBP to ~150GBP. As average length of stay, LTV remained constant over time (apart from increase after the price change). OC&Gym has experienced a gradual drop in the build-up to the price change (caused by increased churn), but this was immediately offset in 2016.

1. C. 2. Customer segments

The goal of this part is to further investigate the differences in LTV across different subscriber segments. Due to the limited scope of this analysis, the most important differences have been chosen and will be presented in more detail below:

There is a large difference between cohorts when split by membership types. This is expected as the subscriber that opts for a flexible membership is by nature more likely to churn. It is however, both memberships seem to have a relatively constant average length of stay.

For the following graph it is important to remember the distribution of the subscribers compared to the total subscriber base. The graph below shows with increasing number of visits, the subscription length increases equally. What does this mean? It shows clearly that the more engaged a subscriber is the more likely he is to be a “long-term” subscriber for OC&Gym. This provides interesting insights that can be used to increase LTV.

1. C. 3. Gyms

Next, we are doing the same for OC&Gyms clubs with the goal of understanding important differences across gyms. We chose the top 3 and bot tom 3 clubs in term on average length of subscription. The best 3 are : Havering, Harrow and Westminster while the botton 3 are : Merton, Brent and Sutton (Redbridge and Hillingdon were ignored as they are too recent and would not allow good comparison over time).

The graph below shows that the average length of subscription varies considerably across different gyms. Even though the graph shows newer gyms having a lower length of subscription, this pattern does not appear when running the analysis for all clubs.

2. What else drives churn other than price? (What other factors influence the likelihood of a sub leaving?

2. A. Hypotheses: “Full gyms during peak hours leads to higher churn”

In order to investigate the hypothesis above, the distribution of gym visits ot the total subscriber base has been compared with the distribution of gym visits of churned subscribers. Whereas it would make sense for subscribers to show a higher churn rate if gyms were full during peak hours, we see only a very weak relationship in the data. Therefore, we can conclude that the capacity of gyms during peak hours does not drive churn.

2. B. Hypotheses: “A change in customer profile over time leads to a higher churn rate”

The graph below shows the average length of stay of subscribers that joined within the first year of a gym opening its doors (group 1) and the average length of stay of subscribers that joined after the first year of a gym opening its doors (group 2). Interestingly, the table shows a positive difference across all gyms, meaning that group 1 stayed longer than group 2. On average the difference is +10% for group one, but in certain cases the difference is as large as 40% (e.g Hammersmith).

We believe that these findings indicate that there is a marginal difference in the make-up of the cohort of subscribers that joins within the first year and the remaining cohort. There is a variety of explanations for this occurrence, but we believe that the findings above indicate a decrease in quality (quality = length of average stay per subscriber) over time. This means that a natural increase in churn has to be accepted to a certain extent.

2. C. Hypotheses: “Churn rate is affected by gym location”

To further analyse this hypothesis, let us remember our findings from 1. A. 3. b). The two gyms, Brent and Southwark, chosen based on their subscriber growth rate in 2018, one bad, one good, show a substantial difference in churn rate. This leads us to the conclusion that the churn rate depends on the location of the gym.

2. D. Other reasons for churn not shown in data

It is important to notice at this point that not all churn reasons can be found in the dataset provided. External factors and market dynamics such as competitor behaviour or industry trends (e.g. more homeworksouts) are therefore, not possible to be assessed. This leaves us with an incomplete picture of churn reasons which shows that OC&Gym should focus on gathering more datapoints on churn behaviour in order to better understand churn reasons

3. What is our benchmark

As for every business, it is important for OC&Gym to have a good unders tanding of how it’s current performance compares to the industry, but also how performance differs across gyms. For this purpose, two benchmark indicators have been chosen:

3. A. Comparison of average growth rate of OC&Gym and industry growth rate

According to IBISWorld (Reference 1) the average growth rate of the UK gym industry has been between 4-5% between the years 2015-2018. This has been compared to the the growth rate of OC&Gym in the graph below. The graph shows that OC&Gym has managed to outperform the average industry growth consistently. However, it must be noted that most of this growth has come from opening new gyms. When we previously looked at growth rate per gym, it becames apparent that many gyms are performing way below benchmark. The goal of OC&Gym should therefore be to establish a natural (excluding growth by new gym openings) growth rate of 5%.

3. B. Lowest churn rate of gym and compared to other gyms

Next to subscriber base growth rate, churn plays a key role in the success of any subscribtion business model. As we were unable to find any information on industry churn rates, we used the best (lowest) churn rate of all OC&Gyms. In the graph below, this has been compared to the churn rate of the total subscriber base. Whereas it is ambitious to establish the same churn rate as in Westminster for every other gym, the graph shows clear room for improvement in churn rate. OC&Gym should aim at reducing churn which will directly lead to an increase in LTV and therefore revenue.

4. What does this suggest as possible options going forward? How could they improve their pricing strategy?

4. A. Standard Annual Membership

Goal: Increase in subscriber LTV

Pricing: 270 GBP (~10x monthly standard membership price, 17% discount compared to a 1-year standard membership)

Impact:

Since we are also suggesting the introduction of an off-peak membership, the standard annual membership will mainly target the peak-subscriber which make up 2/3 of the subscriber base. Of these subscribers, we know that based on past data, that 36% tend to stay between 5 and 10 months. These are the subscribers that we can reasonably expect to switch to an annual membership. As the length of the subscription is unknown when a subscriber first joins OC&Gym, makes us believe that we can use the discount (~17%) of the standard annual membership to boost the sale of annual memberships. We believe that up to 50% of subscribers of the target group above would switch to an annual membership. The increase in expected LTV of 150 to 270 (price of annual membership), is the gain in revenue that can be reasonably expected. The % of subscribers that will switch to an annual membership but would initially have contributed more revenue via the standard membership account for 9.5% of the subscriber base (subscribers that stay 10 or more months). Moreover, we believe that the annual membership can lead to an increase in subscriber acquisition as OC&Gym is able to provide a new offering to its subscribers.

4. B. Off-Peak membership

Goal: Better use of off-peak capacities, Increase of LTV, Acquistion of new subscribers

Pricing: 240 GBP (~9x monthly standard membership price, 23% discount compared to a 1-year standard membership)

Another of the changes to the pricing strategy of OC&Gym that we are proposing is the introduction of an Off-Peak Annual Membership. Similar to the standard annual membership above, the main goal of this new offering is the increase in LTV of off-peak subscribers that make up approximately 1/3 of the subscriber base. As we can see in the graph below, most visits occur during peak-hours, leaving unused gym space in off-peak hours. Therefore, we are proposing a off-peak annual membership for the following time slots (5-7am, 9-12am, 13-17pm, 9-12pm). This does not overlap with the current definition of “peak-hours” that OC&Gym is using, but we believe that this definition should be updated based on our findings.

Impact: Increase in LTV and subsequent increase in Revenue

As mentioned above, a third of our subscribers are visitng the gym mainly in off-peak hours. These subscriber are the target of the off-peak annual membership. Furthermore, we know that 34% of these subscribers on average between 5 to 9 months. Obviously, a subscriber does not know how long he or she is gonna stay with OC&Gym when sigining the inital contract, but we believe this is a good proxy for subscribers that are likely to switch to the Off-Peak Annual Membership (We further discount this number by 50% as not all subscribers will switch membership). The expected increase in LTV is the difference of the current LTV (~150 GBP) to 240, which is the price of the OP-membership. It is important to note that we are expected to lose revenue on the subscribers that would have stayed longer than the 9 months (needed to make up for the 240 GBP). However, this subscriber group only makes up ~13.5% of the total subscriber base. Finally, we expect the OP-membership to furhter drive subscriber acquisition.

4. C. Increasing the price of flexible and standard membership

An additional way to increase the current revenue of OC&Gym would be to increase the price of the standard and flexible membership even further. The analysis above shows that the reaction to the price increase at the end of 2015, even though strong, had only a short-term impact and subscriber numbers and revenue quickly recovered. This could lead to the possible conclusion that a further price increase would create a similar situation and OC&Gym would be able to benefit from the higher prices.

Even though the argument above does carry some weight, we firmly believe that increasing the price of the current memberships is not the correct way to address the current problem. We believe that it does not tackle the underlying issues such a low LTV and a high churn rate. Furthermore, a second price increase within a short time-period could lead to an even further increase in churn. Finally, increasing the price of flexible and standard membership would put OC&Gym closer to the pricing of its main competitor and would therefore lose a key differentiation aspect.

5. Reducing churn by increasing customer engagement

5. A. Reducing churn

In the analysis above, we have been able to see that churn plays an important role in the success of OC&Gym. In order to ensure long-term growth for the company, the churn rate must be reduced as a high-churn rate will not always be able to be compensated by a high number of new subscribers. Especially in areas where the gym market is becoming more and more saturated, this will prove a key challenge for OC&Gym.

We have further seen that churn has different drivers, some more easily to be identified and some more complex. The analysis has clearly shown that churn needs further investigation and more aspects of subscriber behaviour need to be looked into to get a better understanding of why a subscriber is leaving OC&Gym. However, we were able to identify key differences in in churn behaviour when comparing subscriber segments differentiated by the frequency of their gym visits. Subscriber with an average of 2-3 visits per week show a much larger LTV compared to subscribers that visits 1-2 per week. LTV further increases with more visits per week. This shows that the churn rate of more engaged subscribers is clearly lower.

5. B. Collecting further infromation on churn behaviour

Returning to the argument made above, this analysis has managed to shed lights on important reasons why a subscriber is leaving OC&Gym. On the one hand, we have learned a natural increase in a churn rate over time is to be expected as the quality of subscriber decreases with the saturation of the market. On the other hand, we have also seen that factors such as a full gym during peak hours does not lead an increase in churn, even though one might quite reasonably assume. Furthermore, we have seen considerable differences in churn across gyms.

In conclusion, we can say that many different factors need to be considered to better understand churn and consequently develop appropriate measures to reduce it. Therefore, we believe that more information on churn behaviour needs to be collected (e.g survey on churn reasons). This will help to build a more accurate model for churn prediction which is key for a sustainable growth rate for OC&Gym.

Conclusion

The main business problem facing OC&Gym is their slowing revenue growth. This is directly as a result of their slowing subscriber growth rate. The increase in price across subscription types has failed to negate this slowdown. We have found that the company could potentially solve these problems by focussing on increasing the LTV of their existing subscribers through providing an annual subscription and an annual off-peak subscription. As we are pricing these subscriptions above the LTV of our standard subscribers and standard off-peak subscribers respectively, these strategies should provide an uplift to revenue. We believe the low monthly cost of these strategies will also prove helpful in attracting new subscribers to the gyms. Furthermore, past data show a strong correlation between subscriber engagement (as visits per week) and the average length of stay. Therefore, OC&Gym should focus on increasing subscriber engagement.

This report does not provide the evidence to prove our initial hypotheses that there would be a change in the makeup of our subscriber base. In general, it can be said that the different segments analysed are homogeneous and differences are difficult to identify. The major differences and insights have been used to derive the strategic recommendations of this report.

With regards to churn, this analysis has clearly shown that more data on churn behaviour needs to be collected. Combined with the fact that OC&Gym is facing a constant churn rate of 20%, this should become a major focus for the company in the coming months. We firmly believe that through the collection of more data, an accurate prediction model can be built. This can then function as the basis for the development of targeted churn prevention measures.

References

Reference 1 (2020, October 30) Gyms & Fitness Centres in the UK https://www.ibisworld.com/united-kingdom/market-research-reports/gyms-fitness-centres-industry/

Appendix

For subscribers who have the same Join and End date, when did they last visit the gym.

Frequency of active customers that were not covered in 1. A. 1. d) i) Split of Customer Segments

Differences across segments in terms of subscriber join rates and churn rates

Other plots of comparison of High-growth and Low-growth locations

Differences in patterns across segments

Distribution of the avergage subscription duration in months - used in the presentation

Other plots used for context and explanatory analysis

This following cell is a declination of the revenue study included in the report, this time segmented by affluence groups. We created this analysis to understand the impact of subscribers' affluence on OC&Gym revenues.